The strongest single piece of evidence either way sits outside humans and rarely gets cited in these debates. Klimentidis and colleagues (Proceedings of the Royal Society B, 2011) pooled over 20,000 animals from 24 populations of eight species living with or near people, including pet dogs and cats, laboratory primates and rodents, and feral rats, and found mid-life body weight rising over decades in essentially all of them: cats about 10% per decade, dogs about 3%, and lab animals on fixed chow too. That is awkward for pure calories-in explanations of the human trend and is the best argument the contaminant camp has, but it is equally consistent with shared changes in food composition, light and activity, so it does not settle the question, and the field has moved toward ultra-processed-food and GLP-1-pathway explanations rather than a specific contaminant. Which is why the 8% here looks about right: the hypothesis has one good cross-species observation and no identified agent. Dogs sharing our environment is also why they are the species for the first lifespan trials of anti-aging drugs; registry (mine): https://w0lph.github.io/k9/trials/

Canine aging trial registry

Registry of 37 interventional studies and cohorts on aging in dogs, each typed and backed by verbatim source quotes.

How does this resolve if it's found to be a small fraction of what's otherwise multifactorial causation?

The problem is the sort of studies that would point to this unambiguously (RCT under metabolic ward conditions) are no longer performed, and nutrition research is mostly a monument of rubbish as a result. For example, you can still find papers trying to decide at what point between eating 1 egg per day and 2 eggs per day would cause someone to drop dead--all based on p-hacking and dietary recall questionnaires! My personal suspicion is that obesogens are 30% likely to play an important role, so assuming a 1/3 chance over 10 years that the epistemological situation in this scientific community improves sufficiently, that would yield a 10% chance.

I don't trust this random AI app further than I can throw it, but just asking it for fun and it says the results are mixed:

https://consensus.app/results/?q=Is%20the%20chemical%20obesogen%20hypothesis%20true%3F&meter=on

Manifold in the wild: A Tweet by VivaLaPanda

@UrbanLandRent https://manifold.markets/LarsDoucet/will-the-contaminant-hypothesis-of-f27836a908da?referrer=VivaLaPanda

He may not be so good at effective altruism, but prison does wonders for effective physiognomy

Plastic + soy + low-quality food